Distribution network parameter estimation and topology identification are the basis of network planning, operational analysis and security control. The traditional linear regression method has high requirements for measurement data error or noise data, and the estimation is accurate only when there is no noise. However, the actual input measurement values (such as voltage amplitude and phase angle) and output measurement values (such as active and reactive power) have noise data. For topology estimation, even if the measurement error is small, the regression method cannot get accurate topology. Given this, the basic model of distribution network parameter estimation is constructed first, then the influence of measurement error on line parameter estimation and topology identification is quantitatively analyzed. A line parameter estimation model considering the bilateral measurement error is established. It is difficult to analyze the model because of its non-convexity, the minimum Rayleigh entropy problem is obtained by equivalent transformation based on the Lagrange function. Finally, simulation analysis based on the IEEE 8-node system is carried out, and compared with traditional linear regression and the least squares method. This proves that the proposed method has good estimation accuracy even when the measurement error reaches 10%.